An Enhancement of K-means Clustering Algorithm

• Published in: 2009 International Conference on Business Intelligence and Financial Engineering : 24-26 July 2009 pp. 237 - 240
• Main Authors: Jirong Gu, Jieming Zhou, Xianwei Chen
• Format: Conference Proceeding
• Language: English
• Published: IEEE 01.07.2009
• Subjects: Algorithm design and analysis; Analysis of variance; Clustering algorithms; Expectation-maximization algorithms; Gaussian processes; Laboratories; Minimization methods; Monitoring; Partitioning algorithms
• ISBN: 9780769537054; 0769537057
• DOI: 10.1109/BIFE.2009.204

K-means clustering algorithm and one of its enhancements are studied in this paper. Clustering is the classification of objects into different groups, or more precisely, the partitioning of a data set into subsets (clusters), so that the data in each subset (ideally) share some common trait - often proximity according to some defined distance measure. A popular technique for clustering is based on K-means such that the data is partitioned into K clusters. In this method, the number of clusters is predefined and the technique is highly dependent on the initial identification of elements that represent the clusters well. If the numbers of sample data are too large, it may let the cluster members unstable. Another problem is selecting initial seed points because clustering results always depend on initial seed points and partitions. To prevent this problem, refining initial points algorithm is provided, it can reduce execution time and improve solutions for large data by setting the refinement of initial conditions. The experiment results show that refining initial points algorithm is superior to K-means algorithm.